Original Contributions

Transcription

1 15 Original Contributions Epidemiologic Studies of Coronary Heart Disease and Stroke in ese Men Living in, and California: Incidence of Stroke in and Yo TAKEYA, M.D.,* JORDAN S. POPPER, M.D.,t YUKIKO SHIMIZU, D.M.SC.,$ HIROO KATO, M.D.,$ GEORGE G. RHOADS, M.D.,t AND ABRAHAM KAGAN, M.D.t SUMMARY As part of the Ni-Hon-San Study, stroke incidence was compared in the and cohorts. Stroke were classified in two types, intracranial hemorrhage (ICH) and thrombo-embolic stroke (T-E). For each type the incidence in was about three times as great as in. The ratio ICH/T-E was 1/. and 1/1.6 in and, respectively. Blood pressure was the most important risk factor, followed by age for total stroke in both and. Proteinuria was also a risk factor in. Conversely, an index of animal food intake was inversely related to total stroke, significantly in, and at a suggestive level for total and hemorrhagic stroke in. Since the levels of blood pressure do not differ between and, one possible explanation for the large difference in stroke incidence between the two cohorts may be the fact that animal protein and saturated fat intake, which is inversely associated with stroke incidence, is much greater in than in. This explanation would support epidemiologic and experimental studies in which suggest that dietary animal protein and fat exert an inhibitory effect on the incidence of stroke. Stroke Vol 15, No 1, 1984 IT HAS BEEN REPORTED THAT MORTALITY from ischemic heart disease in is low compared with that in the United States and, conversely, that mortality from stroke is higher in.' It is uncertain whether this difference is attributable to genetic factors or environmental factors. In order to examine the relationship of ischemic heart disease and stroke to environmental factors, a large-scale epidemiologic cohort study (Ni-Hon-San Study) was begun in 1965 on males in (Hiroshima and Nagasaki) and ese-american males resident in and in the San Francisco Bay Area of California in whom the difference in genetic factors is minimized and variation in environmental and lifestyle factors is large. The comparative incidence of ischemic heart disease has been reported previously. 34 However, the stroke study was begun later because of delayed participation of the neurologist in, where the prevalence determination was made at the time of examination in Prevalence in in this period and that in and San Francisco at the time of the examination in were compared, and it was reported that stroke prevalence in was three From the 'Institute of Health Sciences, Kyushu University, Fukuoka, ; tthe Honolulu Heart Study, U.S. National Heart, Lung and Blood Institute, Honolulu, ; tthc Department of Epidemiology and Statistics, Radiation Effects Research Foundation, Hiroshima and Nagasaki,, and the SSchool of Public Health, University of, Honolulu,. Address correspondence to: Hiroo Kato, M.D., Department of Epidemiology and Statistics, Radiation Effects Research Foundation, 5- Hijiyama Park, Minami Ward, Hiroshima City 73,. Received October 5, 198; revision accepted August 15, times as high as that in the United States. 5 Subsequent follow-up has now made it possible to compare incidence in for the four years from with that in for the six years from Comparable follow-up for stroke was not carried out in the California study. Subjects and Methods The study subjects are composed of 1,366 males aged years in who were examined during the two years beginning in October 197 and were found not to have stroke at the time of examination and 7,895 ese-american males years in who were examined during the three years beginning in 1965 and were found not to have stroke at the time of examination. Observation periods were four years and six years respectively. Uniformity of diagnostic standards is very important for comparing stroke incidence in different cohorts. In this study, those men who answered "Yes" to the question "Have you ever had a stroke?" and those with positive neurologic findings at the screening examination by the clinic physicians were referred for evaluation by the neurologists (Y.T. in, J.S.P. in Honolulu). For who died between regular examinations, except those that came to autopsy, a detailed questionnaire (Mortality Surveillance Questionnaire) was mailed to the physician who prepared the death certificate (in ), or check was made against the hospital record (in ) for accuracy of causes of death. In order to achieve comparability of stroke diagnosis, the same criteria were carefully followed in both sites. Stroke was diagnosed on the basis of clinical

2 16 STROKE VOL 15, No 1, JANUARY-FEBRUARY 1984 history and neurological findings in the neurologist's consultations which were obtained similarly in both cohorts. Results of laboratory tests or diagnostic procedures in hospitals, such as EEG, cerebral angiography, brain scans, CT scans, etc., were not made available to the neurologists since these tests were rarely used in the cohort. Even in autopsy, the post-mortem findings were eliminated, and the clinical pictures before death were used for diagnosis. This approach insures comparability in those subjects diagnosed by neurologists at the two sites, but eliminates many of the incidence when the diagnoses were based on hospital records and/or laboratory information, particularly in the cohort in which the majority of patients suffering cerebrovascular accidents are hospitalized. Full incidence figures for strokes in the cohort have already been published separately. 6 Stroke were classified by reliability into two grades, definite and possible. Definite stroke was diagnosed in with relatively sudden onset of a neurological deficit which was obvious in the neurologist's consultation or in records of the physician's findings. Duration of the neurological deficit of at least two weeks (or until death) was required. When the exact duration of a neurological deficit was shorter than two weeks but at least 4 hours, or of unknown duration, possible stroke was diagnosed. The study neurologists found no residual signs in most of the latter. Strokes were roughly classified into two sub-types, intracranial hemorrhage (ICH) and thrombo-embolic stroke (T-E). Intracranial hemorrhage combined intracerebral (hemorrhage and subarachnoid hemorrhage because there were which could not be distinguished. When lateralizing signs developed, it could often not be determined whether they were the result of vasospasm or the direct effect of hematoma. Relatively sudden onset of unconsciousness developing during activity, usually with headache, vomiting, nuchal rigidity, fever, and focal neurological deficits, was classified as intracranial hemorrhage. Thrombo-embolic stroke was diagnosed when neurological deficits occurred in the resting or sleeping state, usually without prolonged unconsciousness. More detailed diagnostic classification could frequently not be made because there were where clinical followup or detailed examination could not be conducted. A brief comparison of the results of diagnosis independently made by the neurologists at the two study sites was done by exchanging the same data (6 ). 5 The diagnoses in 1 of these (81%) were in total agreement, and 96.% were in accordance if difference by one grade is accepted. Table 1 shows stroke by reliability of diagnosis and type. The follow-up for death of the study cohorts practically 1% in by checking against compulsory family registration system in and nearly so in and re-examination rate between the two successive clinical examination is very high over, 9%. Thus, person years at risk was computed by multiplying the number of subject at risk and average TABLE 1 Number of Stroke Cases by Subtype and Certainty of Diagnosis and Subtype ICH T-E Unknown Definite Possible 4 6 Definite Possible follow-up period, i.e. four years in and six years in. The age-adjusted average annual incidence rates based on person-year at risk were compared between and and the test was made by Mantel-Haenszel x statistics. Risk factors analyzed in this study include results of physiological tests such as blood pressure at the initial examination, results of laboratory tests such as serum cholesterol and hematocrit levels, and nutrient intake obtained in a 4-hour recall nutrition survey. 7 In the analysis comparisons of means were made after age-adjustment by the direct method except where otherwise specified. Appropriately weighted tests were used. Tests of uniformity and linear increasing tendency were made with the Mantel-Haenszel x statistic by calculating age adjusted incidence by level of each factor. Multiple logistic regression was used for multivariate analyses, the regression coefficients being estimated by the method of maximum likelihood. Absolute coefficients of risk factor in the two cohorts were compared and tested by t-test using standard deviations of absolute coefficient. Results Table shows the average annual incidence (per 1, persons) by age (five-year age groups). The ageadjusted stroke incidence rate (total of definite and possible ) in is 7.4 per 1,, approximately.7 times as high as that in. Examined by age, the incidence in is higher than that in in every age group above 5, and the differences are statistically significant. With the exclusion of possible, the incidence rate in is 3.1 times as high as that in as shown in table 3. In the ratio of ICH to T-E was 1/.; in the ratio was 1/1.6, the number of T-E being larger than that of ICH in both. The incidence rate of intracranial hemorrhage in is 3.1 times as high as that in ; and the incidence rate of T-E strokes in is 3.4 times as high as that in (table 3). However, accurate comparison by type is compromised by the existence of three (4.% of 71) of unknown type in. Means of 15 variables determined at the baseline examination were compared between non-stroke subjects and those who subsequently developed ICH or T- E stroke in order to identify the risk factors for and (table 4). These variables include systolic

3 INCIDENCE OF STROKE IN JAPAN AND HAWAII77a/t<rya et al 17 TABLE Average Annual Incidence of Definite and Possible Stroke per 1 by Age Age Age adjusted ratettt subjects t subjects Testtt tannual Incidence rate is calculated as follows: ; ( /No, of subjects)/4 (Years follow-up). ; ( /No, of subjects)/6 (Years follow-up) tt z test of two rates between two cohorts : p >.1, *: p <.5, **: p <.1, ** : p <.1 tttcalculated by the indirect method with as standard *** *** ** *** and diastolic blood pressure, serum cholesterol, hematocrit, relative body weight, number of cigarettes per day, and nutrient intake per day such as total calories, animal and vegetable protein, saturated and unsaturated fat, simple and complex carbohydrate, and alcohol. Calories per kg of body weight was also included as a possible index of physical activity. In, variables which show significantly higher means for ICH are limited to systolic and diastolic blood pressure. For T-E both blood pressure and cigarette smoking appear to be higher among. The data also suggest an excess relative body weight among the. Conversely, intake of protein, particularly animal protein, and fat, particularly saturated fat, is lower for T-E than for non-stroke subjects. calories and calories/kg also show lower means for T-E. In, both systolic and diastolic blood pressure are again noted to be significant risk factors for ICH. Conversely, serum cholesterol and intake of animal protein, saturated fat and calories/kg are lower for ICH patients. For T-E, systolic and diastolic blood pressure is higher, but intake of protein and fat is lower than that for non-stroke subjects. Next, dose-response for each of the above-mentioned major risk factors was presented as relative risk of incidence (age adjusted) in each level group to the incidence in the lowest level group by classifying TABLE 3 Average Annual Stroke Incidence Per 1 by Age and Subtype Definite Cases Only Intracranial hemorrhagt Age Age adjusted rate stroke by type into ICH and T-E stroke. As shown in figure 1, significant elevation of incidence of ICH and T-E stroke was observed with increase of systolic blood pressure both in and. Negative association was observed between intake of animal protein and fat, especially saturated fat, and incidence of ICH and T-E stroke (figs. and 3). It was particularly remarkable at very low intake group. An association of simple and complex carbohydrate with ICH and T-E stroke was not evident. A consistent association of serum cholesterol with ICH and T-E stroke was not evident in either or in general (fig. 4). However, in the relative risk of ICH and T-E seems high in both extremely high and low serum cholesterol level (U-shape). A negative association was observed between ICH and T-E stroke and calories/kg in, though it was not statistically significant, but no such association was observed in (fig. 5). As some of the risk factors examined by univariate analysis are correlated with each other, a multiple logistic function analysis was performed in order to determine the role of each one of the risk factors while, in effect, holding the others constant. Twelve variables which appeared to be significant risk factors through univariate analysis were chosen and standardized coefficients of risk factors for stroke, i.e., both intracranial Thromboembolic stroke I.I

4 STROKE VOL 15, No 1, JANUARY-FEBRUARY 1984 TABLE 4 Age-adjustedf Means of Selected Variables with and without Stroke Definite and Possible Cases Variable Systolic blood pressure (mm Hg) Diastohc blood pressure (mm Hg) Serum cholesterol (mg/1 m() Hematocrit (%) Relative body weight (%) Cigarette/day Calorie Animal protein (g) Vegetable protein (g) Saturated fat (g) Unsaturated fat (g) Simple carbohydrate (g) Complex carbohydrate (g) Alcohol (g) Calories/body weight (kg) tage adjustment constants ICH 163.4*** 97.** Sug.:p < *: p <.5 **: p <.1 ***: p <.1 hemorrhage and cerebral infarction combined, were estimated by multiple logistic function analysis as shown in table 5. Both in and, the most important risk factor was blood pressure. Age was also significantly related to the incidence of total stroke, while the relation of proteinuria to total stroke was suggestive in, significant in. Conversely, the negative coefficient of animal protein intake was suggestive in, significant in and that of serum cholesterol was negative but not significantly so in either T-E 161.4*** 96.*** *** 39 Sug. 33.* * ** Nonstroke Nonstroke ICH 148.9** 9.5*** 7.4 Sug ** *** T-E 154.1*** 88.8*** ** * cohort. Coefficients of calorie intake per unit of body weight were negative in both cohorts. Neither alcohol intake nor cigarette smoking showed a relation to the incidence of total stroke. ECG evidence of LVH was significantly related to stroke incidence only in. No significant difference between and was observed in the risk factor coefficients except for ECG. That is, there was no evidence of a difference between sites in the magnitude of effects of the risk factors on incidence of stroke. -JAPAN.HAWAII FIGURE 1. Relative risk of intracranial hemorrhage and thromboembolic stroke by systolic blood pressure and country. < S 1SS Sarota Distribution (%> JAPAN HAWAII «H <1 1O SYSTOLIC BLOOD PRESSURE (mmhgl Statistical tait for linear trand J3 *P<OOS Sug 8<P< J5 P>.1

5 INCIDENCE OF STROKE IN JAPAN AND HAWAII/Tatevn el al 19 FIGURE. Relative risk of intracranial hemorrhage and thromboembolic stroke by intake of animal protein and country. < Svnpto DhtrUjutKxi (%) JAPAN HAWAII So». ANIMAL-PROTEIN (gl * Animal protein and saturated fat were combined in an index of animal food intake as a single variable and multiple logistic function analysis was performed with other independent significant risk factors i.e., systolic blood pressure, calories/kg and age for total stroke, intracranial hemorrhage and thromboembolic stroke (table 6). The negative association between the dietary variable and total stroke was significant for, suggestive for. The coefficients for the relation to T-E stroke and to ICH were in the same direction, but statistically suggestive only for ICH in. As indicated, no difference was demonstrated in the kinds of risk factors and magnitude of effects between and. Therefore, the levels of factors in and were compared. Table 7 shows ageadjusted mean levels of each factor in and. There was no difference in blood pressure between and, but hematocrit, serum cholesterol, blood sugar one hour after a 5 g glucose load, and uric acid showed higher levels in than in. Body weight and skinfold thickness also showed higher levels in than in. There were also substantial dietary differences. Intake of total calories, protein (especially animal protein) and fat (especially saturated fat) was larger in than in. While simple carbohydrate level was higher in, complex carbohydrate level was higher in. The intake of sodium was of interest in relation to blood pressure, but it was not possible in this large-scale epidemiological study to estimate accurately the amount of all dietary sodium. Accordingly, estimation was made taking as an index the amount of sodium contained in the components of foods (excluding sodium in salt and shoyu [soy sauce] added in cooking), and it was found that the mean intake in was 4.6 g per day compared with.9 g in. 7 Discussion The ese subjects of the Ni-Hon-San study reside in Hiroshima and Nagasaki. There are remarkable regional differences in stroke mortality in, it being high in the northeastern region and low in the southwestern region including Hiroshima and Nagasaki. Therefore, there is a question whether or not Hiroshima and Nagasaki can be regarded as being representative of. However, the distribution of the prefectures of origin of ese-americans in reveals that those with origin in Hiroshima Prefecture and adjacent Yamaguchi Prefecture account for 4 percent of the migrants and 9 percent have their origin in the southwestern region as a whole. This is very convenient for the comparison of ese in (Hiroshima and Nagasaki) and the ese-ameri- FIGURE 3. Relative risk of intracranial hemorrhage and thromboembolic stroke by intake of saturated fat and country. < Sampte DturltKJtlon %) JAPAN HAWAII S SO- TS <5 SATURATED FAT (gl o.e sore

6 STROKE VOL 15, No 1, JANUARY-FEBRUARY 1984 FIGURE 4. Relative risk of intracranial hemorrhage and thromboembolic stroke by cholesterol and country. < Swnpto Dhtribution l%) JAPAN 1 HAWAII J 4-4S < CHOLESTEROL <m»/c« cans in since this fortuitous circumstance makes it likely that the cohorts under study are genetically similar. It has already been reported that both mortality and prevalence of stroke are higher in the ese than in ese-americans. 15 The present study has found that stroke incidence is approximately three times as high in as in. In both and the United States there has been a fall in stroke mortality since 1968, steeper since 197 and more marked in than in the United States. There are no good nation-wide incidence data in either country. If the incidence data paralleled the mortality data, die effect of the secular trends during the time periods covered in this paper would have been to narrow any difference found between the and Honolulu incidence findings. Since there is little difference between these two cohorts in blood pressure, the most important risk factor, the marked difference in stroke incidence is presumably attributable to other risk factors. Possibilities that come to mind are the differences in the serum cholesterol level; in the amount of cigarette smoking; in alcohol intake; the intake of sodium; and the intake of protein, especially animal protein; and of fat, especially saturated fat. The role of tobacco in the genesis of stroke has been controversial. During the three large prospective studies of cigarette smoking 8 " 1 only one reported an asso- ciation with stroke death in men. 8 In addition Hammond reported an association in women. 1 At Framingham an association was found only for thrombo-embolic stroke in men." In data relating cigarette smoking to stroke are also mixed. Some prospective studies were reviewed by Nakayama. 1 In the largest of these, smoking was associated only with hemorrhagic strokes in women. There was an association with cerebral infarction in one of three smaller studies reviewed. More recent data from a rural Shikoku community were also negative for a smoking effect. l3 In the present data smoking was a significant risk factor only for thrombo-embolic stroke in (table 4). In view of the inconsistency of this finding and the modest difference in smoking habits between and, it seems unlikely that smoking accounts for the difference in stroke frequency between the migrant and indigenous ese. Previous reports from the Ni-Hon-San Study have indicated a larger consumption of alcohol in the cohort than in the Honolulu cohort. 14 Greater alcohol intake could partly explain the greater incidence of stroke in the cohort acting indirectly through its effect on blood pressure ' 6 Alternatively, alcohol intake could enhance bleeding in cerebrovascular lesions by its effect on platelet number and function, 17 or it could, through nutritional deficiency or some unknown mechanism, increase the vulnerability of the TEST FIGURE 5. Relative risk of intracranial hemorrhage and thromboembolic stroke by calories/body weight (kg) and country. SanvMC JAPAN HAWAII < lutrlbutxxi (%) <J CALORY / WEIGHT Ical/kg)

7 INCIDENCE OF STROKE IN JAPAN AND HAWAU/r«/tevo et al 1 cerebral vasculature. A direct relation of alcohol intake to the incidence of stroke, independent of blood pressure, was found for intracranial hemorrhage but not for thrombo-embolic stroke in the cohort. 6 In a stroke survey of residents in Akita and Osaka Komachi et al found a negative association of serum cholesterol level and intake of fat and protein with the incidence of stroke. They suggested that improve- TABLE 5 Standardized Multiple Logistic Function Coefficients Definite and Possible Stroke Cases Standardized coefficient Absolute coefficient Variable Systolic blood pressure (mm Hg) Serum cholesterol (mg/1 ml) Relative body weight (%) Hematocrit (%) Proteinuria.73* * * LVH in ECG Alcohol (g) Saturated fat (g) Animal protein (g) Cigarettes per day Calories/body weight (kg) Age.4 Sug..19 Sug Sug *.6* * *.3* ttests for difference of absolute coefficient between two countries : p >.1 Sug. p <.1 *: p < Testt * TABLE 6 Standardized Multiple Logistic Function Coefficient Definite and Possible Stroke Cases 1CH T-E stroke Variable Systolic blood pressure (mm Hg).74*.51*.65 *.5*.74*.49* Animal protein + saturated fat (g) Calories/body weight (kg) -34 Sug. -.34* -.58 Sug Age Sug.: p <.1 *: p < Sug..6 3* * TABLE 7 Age-adjusted Means of Physical Characteristics and Laboratory Findings and Nutrients Intake Variables Physical characteristics and laboratory findings Nutrients intake Systolic blood presure (mm Hg) Diastolic blood pressure (mm Hg) Hematocrit (%) Serum cholesterol (mg/1 mt) Height (cm) Weight (kg) Back skinfold (mm) Calories Animal protein (g) Vegetable protein (g) Saturated fat (g) Unsaturated fat (g) Simple carbohydrate (g) Complex carbohydrate (g) Alcohol (g) Calories/body weight (kg) *

8 STROKE VOL 15, No 1, JANUARY-FEBRUARY 1984 HYPERTEION CHOLESTEROL HYPEHTEMSION CHOLESTEROL SMALL VESSEL SCLEROSIS Risk (blood pressure, cholesterol) of cerebral in- FIGURE 6. farction. CEREBRAL INFARCTION Not* NfMtrvi rttstlon Pt»tiv rttmrton. mtxtgrsn Irtistnn rrtk ZOI (J) PtMJti* rtistion. hv Irdttlm rrtk 4-6 ) ATHEROSCLEROSIS OF CIRCLE WILLIS ment of nutrition is required for the prevention of stroke (especially in rural areas). In an experiment using stroke prone spontaneously hypertensive rats (SHRSP), 19 an association between stroke and nutrient intake has been observed such that mortality from stroke decreased with the intake of protein and fat. Both blood pressure and stroke incidence increased when salt intake was increased, but when protein or fat intake was increased at the same time, there was increased excretion of sodium into the urine, and stroke incidence decreased. Pathological studies on cerebrovascular lesions reveal small vessel sclerosis and atherosclerosis of the circle of Willis to be the two causes of cerebral infarction, but the frequency of small vessel sclerosis was higher in while atherosclerosis of the circle of Willis was more severe in. Figure 6 is a schematic diagram of the results of the analysis of this clinico-pathological study 1 of autopsy. Small vessel sclerosis appears to be more closely related to cerebral infarction than atherosclerosis. It may be for this reason that the incidence of cerebral infarction is higher in than in. Hypertension is a cause of both small vessel sclerosis and atherosclerosis of the circle of Willis, and consequently it is the most important risk factor for cerebral infarction. However, since there is no difference in the blood pressure level between the and cohorts, factors other than blood pressure need to be considered as the cause for the higher frequency of thrombo-embolic stroke in. Elevated serum cholesterol level may be considered to be the cause for atherosclerosis of the circle Willis but it either has no relation or an opposite effect on small vessel sclerosis. In where the serum cholesterol level is comparatively high, the frequency of small vessel sclerosis tends to be low. The nutrition level reflected by the serum cholesterol value may act on vascular wall structure by suppressing the onset and development of small vessel sclerosis and this may have resulted in the lower frequency of thrombo-embolic stroke in. Although not established because of the small number of of intracranial hemorrhage, the relationship to small vessel sclerosis appears strong. This may also account for the somewhat higher incidence of intracranial hemorrhage in where small vessel sclerosis is more frequent than in. However, the schema shown in Figure 6 must be considered preliminary, requiring substantiation with the accumulation of a larger number of. Kuller and Reisler reported the similar inference from the standpoint of geographical pathology of stroke in relation to the level of blood lipid levels and blood pressure. The criteria for diagnosis of stroke used in the present analysis may seem weak in the light of the recent progress in diagnostic methods, such as development of CT scans, cerebral angiography and so on. This limitation was unavoidable, given the decision to base the study on cohorts over whose medical management the investigators had no control. In the cohort, the collected laboratory findings and diagnostic test results such as angiography were not consulted in making the diagnosis for the present analysis in order to ensure consistency with the ese cohort, for which such information was not completely available. Analysis of the incidence rates and the related risk factors in the same cohort using the full material including laboratory findings and diagnostic tests results was published elsewhere. 6 The effects on the data of restricting the diagnostic information were to reduce the overall stroke incidence from to 16, to decrease the number of definite from 133 to 71, to increase the number of possible from 49 to 55, to increase the number of attributable to unknown from 15 to and increase the ratio of definite ICH/TE stroke from 1/.8 to 1/1.6. The principal effects on the risk relationships of reducing the number of in the current study were to change the alcohol:ich relation from significant to non-significant and the cholestero- 1:ICH relation from significant (p <.5) to suggestive (.5 < p < A). References 1. Gordon T. Further mortality experience among ese Americans. Public Hlth Rep 8: , Kagan A, Harris B, et al. Epidemiologic studies of coronary heart disease and stroke in ese men living in, and California: demographic, physical, dietary and biochemical characteristics. J Chronic Dis 7: 345, Robertson TL, Kato H, et al: Epidemiologic studies of coronary heart disease and stroke in ese men living in, and California: Incidence of myocardial infarction and coronary heart disease. Am J Cardiol 39: 39-43, Robertson TL, Kato H, et al: Epidemiologic studies of coronary heart disease and stroke in ese men living in, and California- Coronary Heart disease risk factors in and. Am J Cardiol 39: 44-49, Kagan A, et al: Epidemiologic studies of coronary heart disease and stroke in ese men living in, and California: Prevalence of stroke. Cerebrovascular diseases, 1th Princeton Conference, edited by P. Scheinberg. Raven Press, New York, 67-77, Kagan A, Popper JS, et al: Factors related to stroke incidence in ese men The Honolulu Study. Stroke 11: 14-1, Tillotson J, Kato H, et al: Epidemiologic studies of coronary heart disease and stroke in ese men living in, and California: methodology for comparison of diet. Am J Clin Nutr 6: 177, Kahn HA: The Dom study of smoking and mortality among U.S. veterans. Report on eight and one-half years of observation. U.S.

9 INCIDENCE OF STROKE IN JAPAN AND HAWAII/To/teya et al 3 Natl Cancer Inst Monogr 19: 1-16, Doll R, Hill AB: Mortality in relation to smoking: Ten years observations of British doctors. Brit Med J 1: , Hammond EC: Smoking in relation to death rates of one million 16. men and women. Natl Cancer Inst Monogr 19: 17-4, Shurtleff D: Some characteristics related to the incidence of cardiovascular disease and death: Framingham Study, year follow-up 17. in The Framingham Study. An Epidemiological Investigation of Cardiovascular Disease. Section 3. Edited by WB Kannel and T. Gordon. U.S. Department of Health Education and Welfare Publication No. (NIH) , Nakayama Y: Epidemiological research in on smoking and 19. cardiovascular diseases. In Atherosclerosis IV. Proceedings of the Fourth International Symposium, edited by CA Schettler, Y Gotto, Y Hata, G Klose. Springer-Verlag, New York Tanaka H, Ueda Y, Hayashi M et al: Risk factors for cerebral hemorrhage and cerebral infarction in a ese rural community. Stroke 13: 6-73, Kato H, Tillotson J, Nichaman MZ, Rhoads GG, Hamilton HB: Epidemiologic studies of coronary heart disease and stroke in Japa-. nese men living in, and California: serum lipids and diet. Am J Epidemiol 97: , 1973 Klatsky AL, Friedman GD, Siegelaub AB, Gerard MJ: Alcohol consumption and blood pressure. Kaiser-Permanente Multiphasic Health Examination Data. N Engl J Med 96: 1194, 1977 Kagan A, Yano K, Rhoads G, McGee D: Alcohol and cardiovascular disease: the an experience. Circulation 64 (suppl III): 7-31, Haul MJ, Cowan DG: The effect of ethanol on hemostatic properties of human blood platelets. Am J Med 56:, 1974 Komachi Y, Iida M, et al: Geographic and occupational comparisons of risk factors in cardiovascular disease in. ese Circulation Journal 35: 9, 1971 Yamori Y, Hamashima Y, et al: Pathogenesis of acute arterial fat deposition in spontaneously hypertensive rats ese Circulation Journal 39: 61, 1975 Mitsuyama Y, Thompson LR, et al' An autopsy study of cerebrovascular disease in ese men who lived in Hiroshima, and Honolulu,. Stroke 1: , 1979 Odoroff CL, Lee KK, Shimizu Y: Preliminary analysis on clinico pathological study on cerebral lesions Unpublished Kuller L, Reisler DM: An explanation for variations in distribution of stroke and arteno-sclerotic heart disease among populations and racial groups. Am J Epidemiol 93: 1-9, 1971 Risk Factors Related to Ischemic and Hemorrhagic Cerebrovascular Disease at Autopsy: The Honolulu Heart Study G. N. STEMMERMANN, M.D.,*T. HAYASHI, M.D.,* J. A. RESCH, M.D.,t C. S.CHUNG, PH.D.,:): D. M. REED, M.D., AND G. G. RHOADS, M.D. SUMMARY This study of cerebrovascular lesions at autopsy among an ese men identifies similar risks factors for cerebral infarction and hemorrhage that have been identified in a previous incidence study. Demonstrated differences were essentially the same whether subjects with these tissue changes were compared to men showing no central nervous system disease at autopsy or when they were compared with men still alive. Cerebral infarcts accompanied myocardial infarction (CHD) in 58% of autopsy and were associated with CHD risk factors (high serum cholesterol, hypertension, severe atherosclerosis of the coronary arteries and aorta). These associations did not persist when CHD were removed from the analysis, indicating there were two subsets of men with cerebral infarction. Hypertension was strongly associated with hemorrhagic disease, as were cigarette use and alcohol consumption. Stroke Vol 15, No I, 1984 THERE ARE MANY DISEASES of the cerebral vessels. These include hemorrhage from congenital anomalies and from mycotic aneurysm, hemorrhage resulting from coagulation defects, bleeding from traumatized vessels, thrombo-embolism, subarachnoid hemorrhage from aneurysm, and intracerebral hemorrhage. The term cerebrovascular disease (CVD) is usually understood to include the last three of these entities and, until recently, has been the most common From the Department of Pathology, Kuakmi Medical Center, Honolulu, ;* The Department of Neurology, University of Minnesota School of Medicine, Minneapolis, Minnesota;t School of Public Health, University of, Honolulu, ;! National Heart, Lung and Blood Institute, Honolulu, ;$ National Institute of ChHd Health and Human Development, Bethesda, Maryland. This project was funded by NIH Contract NOI-HV-91, NHLBI. Address correspondence to: Dr G. N. Stemmermann, Kuakini Medical Center, 347 N. Street, Honolulu, Received July 6, 1983; revision accepted September 14, cause of death in. 1 In recent years the mortality from CVD in has been on the decrease and may soon be superseded by cancer. In an ese CVD has been displaced to third place after malignant disease and heart disease, a trend that was already apparent by I96. A comparative study of cerebrovascular disease at autopsy in ese men who lived in Hiroshima, and in Honolulu, 3 indicated that infarction was more frequent in Hiroshima than in Honolulu, but that hemorrhagic disease was equally frequent in the two cities. This is a report of the epidemiology of cerebrovascular disease, as seen at autopsy among the an ese men. It supplements prior reports on the epidemiology of clinical cerebrovascular disease in the same population. 4 3 Cerebrovascular disease does not lend itself as readily to epidemiologic study as coronary heart disease, because it is less common, is less frequently fatal and occurs in older persons than does